1
Task Characteristics and HRM Activities on Online Labour Platforms
MSc Business Administration: Human Resource Management 27-08-2021
Student: Jaap Cortjens (s1675117) Committee Chair: Jeroen Meijerink
Committee member UT: Maarten Renkema
Acknowledgements
Firstly I want to thank Jeroen Meijerink for supporting and understanding me throughout the process of writing this difficult thesis in Covid-19 times. Secondly I want to thank Maarten Renkema for providing a fresh set of eyes that helped find the flaws I skipped over after going over the thesis too many times. I would also like to note my appreciation for my girlfriend’s ever positive attitude and unwavering support while writing this thesis. Lastly I want to thank my friends, flat mates, and family who have helped me in any way towards finishing this thesis.
Abstract
Online Labour Platforms (OLPs) where gig workers complete often short tasks for payment are becoming a more common and established form of business where supply and demand of short-term labour are matched by the platform. Even though OLPs do not employ people to complete tasks via their platform, they however do use HRM activities to steer the behaviour of the people using the platform. Different types of OLPs have been identified by scholars as well as different HRM activities commonly used by OLPs. However, there are differences in the used HRM activities among platforms. It is a reasonable assumption that when different tasks are offered via OLPs the HRM activities to best facilitate these tasks are also different.
This study set out to study the relationship between the different job characteristics and the
different corresponding HRM activities by interviewing the representatives of 10 diverse OLPs
operating in the Netherlands and analysing their responses, as well as conducting a netnography
by signing up to those OLPs as a gig worker and as a client. As a result the task characteristics
did have an effect on the HRM activities through the dependence they created between clients,
gig workers, and OLPs. However, other factors such as the industry and life cycle stage of the
OLP were also impactful on the HRM activities implemented by the OLPs.
2
Contents
Work Characteristics and HRM Activities on Online Labour Platforms ... 1
Acknowledgements ... 1
Abstract ... 1
1. Introduction ... 4
2. Theory ... 6
2.1. Online Labour platform typologies ... 6
2.2. HRM Activities... 11
2.3. Resource dependence theory ... 17
2.4. Propositions ... 20
3. Methodology ... 23
3.1. Data Collection ... 23
3.2. Operationalization ... 24
3.3. Data Analysis ... 24
4. Results ... 26
4.1. General data set information... 26
4.2. HRM Content ... 27
4.3. Client and Gig Worker Matching ... 28
4.4. Recruitment Activities ... 31
4.5. Selection activities ... 34
4.6. Training and Development ... 35
4.7. Compensation activities ... 37
4.8. Job Appraisal ... 39
5. Discussion & conclusion... 42
5.1. Discussion and Implications for theory ... 42
5.2. Implications for practice ... 44
5.3. Limitations ... 44
5.4. Recommendations for future research ... 45
5.5. Conclusion ... 47
6. Acknowledgements ... 47
7. References ... 48
8. Appendices ... 55
8.1. Appendix 1: Detailed OLP typology retrieved from Schmidt (2017) ... 55
8.2. Appendix 2: Interview protocol (Dutch) OLP representative. ... 56
8.3. Appendix 3: Detailed interview structure ... 61
3
8.4. Appendix 4: List of a priori codes ... 63
8.5. Appendix 5: Power relations between actors on the interviewed platforms ... 65
8.6. Appendix 6: Overview of practices per actor per HRM domain per platform ... 67
8.7. Appendix 7: Elaborate philosophy per platform ... 76
4 1. Introduction
Almost everywhere on the world you can nowadays summon a cab via a mobile app, while hiring a web-developer on the other side of the world to build a website for you. Online labour platforms make it available for anyone to buy all kinds of goods and services at any time.
Taylor and colleagues (2017) simply described this development of the workforce as “people using apps to sell their labour” (p. 25). But the labour platforms themselves that enable people to sell their labour digitally, are only possible because of the technological developments of the last decades (Gandini, 2018). These platforms connect labourers and consumers instantaneously, as a service, but also as a digital marketplace (Veen, Barrat & Goods, 2019).
These platforms have taken the world by storm, are everywhere, and upset the established workforce (Kuhn & Maleki, 2017). Because of its huge impact, and its disruptive (De Stefano, 2016; Kost, Fieseler & Wong, 2019) but also enabling (Kost, Fieseler & wong, 2018) capabilities it is important to understand how these labour platforms work. The use of these platforms is expected to grow annually by 24% (Kässi & Lehdonvirta, 2016), which makes it even more relevant. Besides the platform usage growing, the uses for platforms are also growing. The range of tasks completed through platforms is just as well growing (Duggan, Sherman, Carbery & McDonnell, 2019), as well as that tasks are created that did not exist before (Wong, Fieseler & Kost, 2020).
The conditions under which platform workers have to work, what challenges they face, and what the gig-worker population looks like is one strand of research already explored (Tran &
Sokas, 2017; Ashford et al., 2018; Kost, Fieseler & Wong, 2019b; Wood, Graham, Lehdonvirta
& Hjorth, 2019; Maurer, Mair & Oberg, in press.). This strand of research helps to better understand the type of people working on these platforms and how this type of working affects them. As a result of the HRM used on these people, the platform workers doing gigs are aligned with the strategic goals of the platform, with no or little contribution to the formulation of these goals (Meijerink & Keegan, 2019; Gegenhuber, Ellmer & Schüßler, 2020). Platforms try to steer their workers into serving the clients or end-users of the platform as well as possible.
To exemplify this, Veen, Barrat and Goods (2019) showed how UberEATS in Australia uses algorithmic HRM to monitor job progress through an app, constrain worker choices through obscuring information, and appraise performance for further work assignment or consequentially fire underperformers.
Studies have also been done more specifically on what online labour platforms do in terms of HRM, to grow and to have platform workers stay with them. This research strand highlights the effects of algorithmic HRM, exposing autonomy paradoxes, workers’ perceptions of fairness, and much more (Lee, Kusbit, Metsky & Dabbish, 2015; Duward, Blohm, Leimeister, 2016; Kost, Fieseler & Wong, 2018; Wu, Zhang, Li, Liu, 2019). While this gives more insight in HRM practices, these studies are mostly focused on the consequences of these HRM practices, but not why they have been implemented in their given context.
Additionally there are studies that typologize all online labour platforms along dimensions that help make sense of the differences and describe this part of the employment economy (Schmidt, 2017). Characteristics such as platform worker control and wages (Kalleberg & Dunn, 2016), or task-structure, dependence on other people, and commitment (Nakatsu, Grossman &
Iacovou, 2014) are used. These typologies show what the differences between the task
characteristics on the platforms are, but these typologies or taxonomies are then used to
describe the differences and not to explain or predicts selected outcomes. This is striking, since
5 differences in types of platform tasks may result in differences in HRM activities adopted by online labour platforms. Based on this previous research, it is thus reasonable to suspect that the characteristics of the tasks offered on online labour platforms such as standardizability of tasks leads to different HRM activities. Therefore this research proposes to look into the effect that the characteristics of the tasks done on an online labour platform have on the implemented HRM activities, asking the following question: “In what way do task characteristics influence the HRM activities practiced on online labour platforms?”
What this research contributes to the literature is firstly a review of the different
characterizations of the tasks completed via online labour platforms, as well as new insights in
the different HRM activities implemented on online labour platforms. Ultimately, this study
contributes to a deeper understanding of how different HRM activities come to be amongst
online labour platforms. To do so the research about HRM in online labour platforms, platform
typologies, and platform worker characteristics need to be combined, and ideally result in
insights in the effect of work characteristics on HRM application in the platform economy.
6 2. Theory
To properly answer the question how the type of work determines the HRM used on online labour platforms, the involved concepts need to be specified and clarified. Firstly the characteristics of the tasks offered via platform are explained to understand how they differentiate from one another. Secondly the way HRM can be studied on online labour platforms is theorized.
2.1.Online Labour platform typologies
As described in the introduction, online labour platforms are defined as being firms that use technology to fill short-term labour needs with independent contractors (Kuhn & Maleki, 2017). Jobs on online labour platform are typically short tasks that firms or individuals do not want to do in-house/themselves (De Stefano, 2016). The online labour platforms do not execute the tasks. Those platforms are a digital market place were supply (independent contractors) and demand (people or companies in need of a good or service) can find each other (Minter 2017).
Online labour platforms are often not a place where supply and demand randomly meet. Most of the time, a key function of successful online labour platforms is appropriately matching of independent workers and clients to one another, and managing that relationship (Gandini, 2018;
Jarrahi & Sutherland, 2018; Lehdonvirta, 2018). This means that on these labour platforms there are at least three actors involved in the transactions conducted.
As stated before, the supply of labour on the online labour platforms consists typically of independent contractors. According to Spreitzer, Cameron and Garret (2017) the status of independent contractors comes from changes in the macro-economic landscape: “Short-term financial results drive decision making, firms seek flexibility through employment at will to meet changing demand” (p. 476). Kalleberg (2009) described this as precarious work, with little protection or security the platform workers are more at the mercy of their employers.
Individual contractors need to arrange their own insurances or take the risk. Besides this change in the macro-economic landscape, technological development also enables this working arrangement, making it possible to work in the cloud, and split up work into individual tasks for anyone to do, reducing the need for fixed employees (Dunn, 2017; Spreitzer et al., 2017;
Lepanjuuri, Wishart & Cornick, 2018).
These individual tasks are typical for online labour platforms. Meijerink and Keegan (2019) described it as “…the sourcing of tasks by a requester (which can be either a firm or an individual consumer), which are relatively short-lived and performed by independent workers who move from one assignment (or ‘gig’) to another.” (p. 6). Not all platforms have short-lived gigs of riding people for 5 minutes like Uber. Upwork for example usually has gigs that last several days to complete, but are still relatively short compared to conventional employment settings (Schmidt, 2017).
Besides the client and the independent contractors, the platform takes an active role in
facilitating transactions occurring over the platform. Next to matching supply and demand,
platforms often also set the tariffs of, set the requirements for, and monitor performance of
work that traffics the platform (Wood et al., 2018; Meijerink & Keegan, 2019). When the
platform controls either or several of these elements it is mediating the transactions, but when
it is only enabling on a technical level, it is an infrastructure provider (Schmidt, 2017). Usually
however, the platform is the mediator (Bonet, Cappelli & Hamori, 2013). Schmidt (2017) also
explains how this three-way relationship leads to a power asymmetry. While the platform has
7 access to all data and information related to the interactions, the supply and demand side only see a small window with limited information. Additionally, the platform can easily and cheaply up- or downscale, while the other actors face more risk in doing so (Frenken, Vaskelainen, Fünfschilling & Piscicelli, 2018).
In conventional employment settings the job characteristics are often studied using the Job Characteristics Model (JCM) by Hackman and Oldham (1976). This model states that work should be designed to satisfy the five core job characteristics of variety, autonomy, feedback, significance and identity in order to enhance the psychological state of employees and improve performance (Parker, Morgeson & Johns, 2017). Although this model is often elaborated on with additional characteristics, the core five characteristics are well established in the job design literature (Parker, 2014). The next section will first elaborate on how the online labour platform literature has differentiated between platform types, and then abstract the common characteristics of tasks. Then these characteristics will be further explained using the OLP literature and the job design literature.
Several characterizations of the platform types have so far been contributed to the online labour platform literature (Cappelli & Keller, 2013; Nakatsu et al., 2014; De Stefano, 2016; Schmidt, 2017; Duggan et al., 2019). These are summarized in Table 1:
Author(s): Platform types:
Cappelli and Keller (2013):
online labour platforms as a branch of “contract work”
1. Subcontracting vendors 2. Sourcing arrangement Nakatsu et al. (2014): a
taxonomy of crowdsourcing - Contractual hiring (structured, independent task) a. Low commitment:
i. Human Intelligence tasks ii. Crowd sharing marketplaces b. High Commitment:
i. Online employment platforms - Distributed problem-solving (structured,
interdependent task) a. Low commitment:
i. Geo-located data collection ii. Distributed knowledge gathering iii. Crowdfunding
- Solo New Idea generation (unstructured individual task)
a. Low commitment:
i. Consumer-driven innovation b. High Commitment:
i. Online problem-solving platforms ii. contests
- Collaboration (Unstructured, interdependent task) a. Low commitment:
i. Real-time Idea jams b. High Commitment:
i. Open source/content development,
design and projects
8 Kalleberg & Dunn (2016):
based on common work types
1. Transportation Platforms 2. Delivery/Home Task Platforms 3. Crowd work Platforms
4. Online Freelance Platforms De Stefano (2016): just-in-
time workforce 1. Crowd work: completing a series of tasks through online platforms
2. Work on-demand via apps: this is when traditional tasks are completed through apps managed by firms that connect supply and demand and intermediate quality standards and prices.
Schmidt (2017): Digital
labour 1. Cloud work (web-based digital labour)
a. freelance marketplaces b. micro-tasking crowd work
c. contest-based creative crowd work 2. Gig work (location-based digital labour)
a. Accommodation
b. transportation and delivery services (gig work)
c. household services and personal services (gig work)
(for a full overview see Appendix 1) Duggan et al. (2019): app-
work (building on Cappelli
& Keller (2013))
1. Capital Platform Work (sell goods or lease assets through platforms)
2. Crowd work (geographically dispersed split up digital labour)
3. App-work (service providing on demand on location)
Table 1: Platform typologies in literature
These typologies have different numbers of categories, and focus on varying elements. When studying these typologies several task characteristics are used across online labour platform literature to distinguish between platform types. Those are abstracted into table 2 as a list of proposed dichotomies of characteristics of task on online labour platforms:
Characteristic dichotomies:
Standardised work Unstandardised work High skill level required Low skill level required
Online offline
On-demand/short task Ongoing process/long task Individually tasked Crowd tasked
High personal investment Low personal investment
Table 2 Proposed work characteristic scheme based on existent literature
The first task characteristic that can be bought and performed through online labour platforms
is whether it is standardizable or not. Kalleberg and Dunn (2016) described this as autonomy
in work. Can the worker determine how to perform a task, or is there a pre-set procedure on
how to do this. To exemplify, micro taskers encounter an explanation of the task when
accepting the task on how to do it, and another way is not allowed (Pichault & Mckeown,
2019). But when an online freelancer is asked to develop a piece of software, the end-goal or
9 result is predetermined, but the process is up to the online freelancer. Nakatsu and colleagues (2014) called this well-structured and unstructured, basing their argumentation on whether there is a predetermined solution. If there is not, inventiveness is required to come to a solution.
The example they gave is the developing of innovative ideas, or create a computer algorithm.
Schmidt (2017) also used the term creativity to define the unstandardised tasks, and spoke of pre-set procedures, and routes to follow in the case of ride-hailing platforms. Logically standardizable tasks result in a more replaceable workforce than ones where the process is less defined (Bonet et al., 2013) This dichotomy uses both the task variety and task autonomy as explained in the JCM, the degree to which a job is repetitive and does not require different skills as well as the gig worker’s lack of discretion about decisions in the work process lead to less fulfilled and motivated gig workers than gigs with tasks with more skill variety and work process discretion for the gig worker (Hackman, 1980).
The second dichotomy is about education or skill level that is required to perform a task.
Although this characteristic is somewhat related to standardizability, they are not the same.
Where standardizability or structuredness relates to the variety or lack thereof in the work, skill level relates to the capacities needed to complete the given task. It is often the case however, that more standardised tasks require less education or skill development (Kost et al., 2018).
Schmidt (2017) gave the example of freelancers being specialists, who because of their education can negotiate higher pay, but micro taskers are generally unskilled and are paid what the requestor determines for the task. Kalleberg & Dunn (2016) added to this that home chores and delivery/transport platforms also offer tasks that do not require any special skills to complete. Thus being more educated in the gig economy, performing tasks that require special skills grants a more powerful position to those gig workers (Kost et al., 2019). Whilst this also relates to the job variety as explained in the JCM, this also relates to the moderator of knowledge and competences that needs to be satisfied to be able to achieve the psychological states and increased performance described by the JCM (Ploher, Noe, Moeller & Fitzgerald, 1985). When a gig worker possesses the necessary skills and knowledge to complete tasks they are more likely to experience positive emotions in their work (Parker, 2014).
Another characteristic so far not mentioned specifically, but presented in almost all literature on the gig economy is whether the work is online or offline. This characteristic refers to whether the job needs to be performed digitally or on a specific geographical location, the point of production as Gandini (2018) puts it. Terminology may vary on this matter. Literature speaks of online – offline, digital – local, real-life – digital, and gig – cloud work (Cappelli & Keller, 2013; Nakatsu et al., 2014; Dunn, 2017; Schmidt, 2017; Duggan et al., 2019). The benefit for requestors of online gig work, is that the pool of potential gig workers is world-wide. For offline work the gig workers need to be geographically nearby the client to be able to perform the task (Rosenblat, 2018). Another element that is more important in offline gig work is that social appearance, or public image, is more important for the success of the platform (Schmidt, 2017).
Rosenblat (2018) explains that this exposure is important for the growth of offline platforms,
and that the gig workers directly impact the reputation of the platform. Therefore the gig
workers’ performance is monitored in more detail. For example, Uber driver’s phone shakiness
during gigs is measured to see how safe the driver drives (Rosenblat, 2018). The online or
offline dichotomy has no importance for the JCM, except for a minor effect on the job
significance characteristic, where the public exposure of the gig worker during offline tasks
may provide a slight feeling of job significance and identification with the OLP.
10 The duration of tasks performed on online labour platforms is the fourth task characteristic investigated. “Jobs on ride-share platforms (Uber and Lyft) are typically less than ten minutes (the average ride is three to four miles long). Jobs from Handy and TaskRabbit typically can be completed within the same day, while jobs on sites like Upwork and Freelancer are commonly project based and tend to have longer durations” (Kalleberg & Dunn, 2016, p. 12).
For the platform this means that their approach to monitor the jobs fulfilled, is dependent on whether the jobs are so short or whether they take longer to complete (Schmidt, 2017). Nakatsu and colleagues (2014) called this characteristic commitment of time by the gig workers, but also uses this term for the sixth of these work characterizations. The JCM in this context focusses on job identity, whether the gig worker completes a job or merely completes a task contributing to a job or project, where the latter is less motivating for a gig worker (Parker, 2014).
Then the one but last characteristic is about who get offered the gig. Is the gig offered to a selected individual, or is the job offered to a crowd of people. Duggan and colleagues (2019) who divided gig work in app-work, capital-platforms, and crowd work said the following: “In app-work, an algorithm quickly identifies and offers labour to one person, whereas in capital- platform work and crowd work, it is the customer or requester who decides and selects whose services to pay for” (p. 8). This signifies whether select individuals are offered the task, or if the task is out there for anyone to take it on (Schmidt, 2017). Besides whether individuals get an offer, or whether it is offered to a lot of workers is one thing, but then there is also the difference between having to complete a task, or if these tasks need to complete by several people, as a virtual team (Nakatsu et al., 2014). For this research however, the focus is on whether individuals are offered a task, or whether it is available for anyone to accept. Because in this case the difference in task assignment is the intended characteristic, and doing jobs alone instead of in teams is most often the case on online labour platforms (Taylor, 2017). Whether the gig worker is chosen for a task or is able to choose the gigs they prefer is also strongly tied to he autonomy characteristic of the JCM where choices such as when to do tasks provide more motivation and performance according to the JCM (Parker, 2014).
Lastly, the investment a gig worker needs to make to start working through a specific platform also impacts accessibility to a platform. for ride-hailing you need a car, while for micro-tasking or delivering by bike the capital needed to start is much lower (Schmidt 2017). Similarly, Duggan et al. (2019) say that for crowd sourcing, a platform worker only needs a stable internet connection, while a ride-hailer needs to have a car, and is exposed to more risk in traffic. This risk is especially influential given the freelance status that these platform workers have.
Nakatsu and colleagues spoke in this case again about commitment, where high-commitment is the case when more resources need to be expended to perform, where low-commitment tasks
“require crowd response, but not intensity” (Duggan et al., 2019, p.832). Doing training/courses or purchasing software can also be seen as resources needed to be expended before being able to perform (De Stefano, 2016). The JCM considers this to be job resources that help to deal with the job demands. Being able to better cope with job demands increases the long-term motivation (Kopelman, 2006).
These characteristics all have in common that they say something about the scarcity of workers
available. For example it is easier to find uneducated people to fulfil tasks than to find a
specialist. It is also easier to find people who do not need to highly invest personally before
11 starting to work. When there is less scarcity, that means that there is wider access to resources.
The significance of this will later be explained in section 2.3.
2.2.HRM Activities 2.2.1. HRM relevance
HRM is of vital importance for the sustained success of a firm, when it comes to creating, improving and maintaining value (Sparrow, Hird, Hesketh & Cooper, 2009). What this value may comprise is widely discussed among HRM scholars (Lepak, Smith, and Taylor, 2017).
But in essence, HRM is practiced to improve the performance of a business in its core functions (Alewell & Hansen, 2012). The business models of online labour platforms differ, so assumingly so do the HRM systems they employ. Johnson (2019) Found that indeed HRM activities differ between platforms. In a traditional sense HRM is conducted by the firm employing a workforce (Lepak & Snell, 1999). But in the platform economy where employment relationships are actively avoided, so is the formal existence of an HRM structure (McKeown, 2016). However, HRM still occurs as noted before, and is especially necessary to commit or control the gig workers. This is because control and commitment cannot be achieved through the employment agreement otherwise in place. Unsuccessful HRM activities can lead to losing clients and/or gig workers to other platforms, or going out of business altogether (Duggan et al., 2019).
The scope of HRM in the platform economy is not as much aimed at facilitating the workforce, but more at facilitating the connection between requestors and gig workers. As Meijerink and Keegan (2019) define this scope: “the multilateral exchange relationships among intermediary platform firms, gig workers, and requesters” (p. 4). Besides the fact that HRM is not focused on the workforce in the gig economy, more dubious differences arise in HRM planning and implementation in the gig economy. the traditional structures do not apply in the gig-economy, creating paradoxical and exiting circumstances. Explaining how this paradoxical HRM circumstance occurs may be best explained using Ostroff and Bowen’s (2000) division between HRM content and HRM process, to cover all bases where HRM differs so tremendously from the traditional setting.
2.2.2. HRM content
The content of HRM activities is what is intentionally employed by the firm to reach its organisational goals and values (Bowen & Ostroff, 2004). Employing isolated HRM activities to direct and facilitate the work force on online platforms (or in any context for that matter) often lead to no, little, or negative results (Jiang et al., 2012). Therefore a consistent bundle, or system usually result in more fruitful performance (Zhou, Hong & Liu, 2013). Employing such a system is based on the HRM philosophy. A firms HR philosophy entails the guiding principles that characterize a firm’s attitude towards employees (Kepes & Delery, 2007).
Moreover Schuler (1992) described HR philosophy as a firm’s statement on how human
resources attribute to success. So philosophies are what guide a firm into employing a specific
HRM system (Monks, Kelly, Conway, Flood, Truss & Hannon, 2013). In OLPs the human
capital is not internal to the organisation, so the attitude towards it is different from that in
conventional organisations to start off with. Gig workers, the human capital, are a much more
uncertain entity than a workforce, and are approached with a different attitude.
12 With the notion of philosophies in mind, Lepak, Liao, Chung and Harden (2006) set out and defined different conceptual HR systems that are to be found in the literature, distinguished by the different philosophies they fundamentally serve. Even though execution of a system might not be the same in different contexts, the conceptual difference is what matters at this strategic level.
HR system for control is about setting clear goals, boundaries, and control in general. In this system, the philosophy values workers as replaceable cogs in the machine, and the impact on the labour process by workers should be minimized (Arthur, 1994). To do so tasks are made as standardised and procedural as possible, and the doing of tasks is separated from the thinking part of the work (Guthrie, 2001). Close performance monitoring is also a part of using a control HR system, in the context of platforms this is highly enabled by the use of technology (e.g.
Uber driver rating system directly influencing how often a job is offered) (Good et al., 2019).
HR system for commitment was at first posed as the only alternative to control systems. The philosophy served by this system is that the workers are valued individuals who should identify and align with the organisational goals by their own choice. The firm attempts to have the employees identify with the organisational goals, so that they work hard to accomplish those goals (Arthur, 1994). To have them commit their effort towards the organisational goals, practices such as intensive training and development, but also high wages and promotion from within are practices seen in such a system (Whitener, 2001). In the platform context internal promotion is not viable since gig workers are not employed, but there are for example on Upwork possibilities to earn rising talent status, top rated status, and access to premium service when freelancers commit more of their effort towards improving their profile and performance (Upwork, 2020).
HR system for employee involvement is somewhat related to commitment systems, but in this system the firm attempts to empower employees through information flows, influence on decision-making, rather than having them commit to the pre-set goals by the firm (Zacharatos, Barling & Iverson, 2005). In this system the philosophy on the workforce is thus one where the human capital also fulfils strategic functions besides the core production functions towards success, human capital is the business. Practices like job rotation, employee problem-solving groups and product innovations thought of and implemented by employees are considered to fit with this system (Osterman, 1994; MacDuffie, 1995). In the platform context such decision- making power with the gig-workers is not common (Schmidt, 2017). Unless the platform is managed by the gig workers, but this tends to be less successful (Sriraman, Bragg & Kulkarni, 2017).
High performance work system (HPWS) is then again built up on commitment and
involvement, focusing on the potential competitive advantage that a firms employees can bring
to the table. This system uses a philosophy of human capital as the most valuable resource of
the company, and a source for potential new competitive advantage. To make use of that
potential workers are treated with respect, and trust between management and workers is
essential (Huselid, 1995). Practices include wide varieties of benefits, individual and group
incentives, work-life balance programs, and intensive training. According to Huselid (1995)
this system attempts to combine a wide range of best practices to retain talent and weed out
under-performers.
13 There were two more systems presented in Lepak and colleagues’ (2006) work on HRM systems. HR system for Occupational Safety and HR system for Customer services are left out of this research, since they are more case specific adaptations of HPWS, but are not philosophically distinct from HPWS. Lepak and colleagues (2006) stress that there is no one best system that fits all organisations. Besides that, only the philosophy of a system stays the same when implemented. Focusing on economic or innovative factors may lead to the same HRM systems, but with entirely different policies and practices comprised within them (Alewell & Hansen, 2012). Looking at the OLPs, none of these systems are a perfect match, since the workforce is not internal, as said before.
To effectuate their HR Philosophy, organisations have HR policies in place (Schuler, 1992).
“They are employee-focused programs that impact choices regarding HR practices” (Jiang et al., p. 75, 2012). According to Wright and Boswell (2002) firms communicate their intentions about HR processes that ought to be exercised in the firm through policies. Policies in turn can lead to practices guided by the policy chosen by the firm. Jiang and colleagues provided an overview of the policy domains existent in literature, which are supplemented with practices observed on OLPs described in table 3 below:
Policy domain Description Example HRM practices
Recruitment Intentions regarding the hiring of employees. Employee characteristics, and employment strategies are included.
Referral schemes, internet advertisement
Selection Intentions regarding appointing work to the members the workforce.
Algorithmic matching,
Featured gigs and/or gig-workers Training and
development
Intentions towards developing the skills of the workforce further (or not)
Fourth party courses, suggest tips and tricks for gig-workers
Performance management
Intentions regarding the appraisal of work, and what is to be done with that judgement
Gig-worker performance rating, Client reputation rating,
Job duration monitoring Compensation Intentions related to the payment for
work done by the workforce
Payment algorithms, price setting Incentive Intentions towards ways of motivating
the workforce
Surge pricing, Referral bonus Involvement Intentions towards involving the
workforce in decision-making
Requesting gig-worker feedback
Job Design Intentions towards what elements (and to what extend) to include in a job/function of the members of the (potential) workforce
Demarcate job boundaries, Gig-worker planning,
Communicate task requirements
Table 3: Policy domains retrieved from Jiang et al. (2012) complimented with example practices seen on OLPs
To execute the policies as described before, the OLPs instigate HRM practices to fulfil them.
Practices are the activities to achieve specific outcomes (Becker & Gerhart, 1996). The practices can be grouped by the policies that they are instigated for, but that does not mean they cannot also contribute to the other policies and OLP values. As the philosophy, system, and policy are communicated concepts of content, HRM practices are the most observable element.
Even though all these policy domains are necessary according to Boxall & Purcell (2008),
Platforms seem to focus more on several of these while neglecting others.
14 Since there is a high through-put rate of gig workers in a platform, finding new gig-workers is essential to match the demand (De Stefano, 2016; Kuhn, 2016). However, most of the time gig workers join platforms by themselves due to a lack of other options (Bellace, 2018; Veen et al., 2019). The recruitment policy of platforms is thus not strongly defined, and rather neglected most of the time. Most recruitment happens through informal networks of gig workers (Wood et al., 2019b). Platforms do have referral schemes in place to further draw in gig workers (Goods et al., 2019), as well as hold digital marketing campaigns (Ashford et al., 2018). On platforms where gig workers approach clients or clients approach gig workers Algorithms are still helpful in bringing gig workers or clients with a good reputation or set of skills to the foreground (Ettlinger, 2017; Schmidt, 2017).
Making sure that clients are adequately served is an OLPs core task (Schmidt, 2017). The platforms’ selection policy is thus one of the more focused policies. Most platforms use the reputation scores comprised to accomplish the performance management policy to match higher reputation gig workers with clients (Rosenblat & Stark, 2016; Jarrahi & Sutherland, 2018). Gegenhuber and colleagues (2020) add to this that platforms usually have a core of gig workers that have the higher reputations and are available most often that are matched to most of the tasks, and the periphery is offered tasks when the core is unavailable. Platforms practice this with algorithms to match clients and gig workers as fast as possible (Lee et al., 2015; Lee, 2018).
From a study by Möhlmann & Zalmanson (2017) about algorithmic management in Uber, it seems that the platform opts to attempt to control gig workers through strict performance management policies, whilst also attempting to have them commit through compensation and incentive policies. Wu and colleagues (2019) showed the same happening in China, but added that incentivization is the most important policy for Uber, using practices such as surge pricing in busy areas, daily bonusses, and peak-earning guarantees. Wood and colleagues (2018) contributed that for online work the performance management is as strict, if not stricter, as Uber, employing practices such as screenshotting and registering average keystrokes while working on a gig, as well as the rating and reputation systems in place. Compensation is again more commitment based, which showed through payment-guarantee if all monitoring practices were enabled, and higher payment for more complex tasks. This paradox in their HR philosophy is hard to get around. On the one hand the platforms view the gig workers as entities that need to be controlled for firm performance, while on the other hand they need to be committed to achieve firm performance.
Training and development is a more neglected policy (Kuhn & Maleki, 2017). This is because training would imply employment relationships. To ensure quality, OLPs advise gig workers to use the fora for gig workers to learn from one another, and/or follow fourth party courses (de Stefano, 2016; Kost et al., 2018; Goods et al., 2019).
Gegenhuber and colleagues (2020) studied the involvement policy in medium sized platforms in Germany. They show that while gig-workers are granted a ‘microphone’ (a way of expressing their attitude towards the platform), they are not granted a ‘megaphone’
(opportunity to broadcast their attitude to everyone). Gig-workers are compelled to stay when
they can express their attitude, but platforms make sure they are not involved with the decision-
making to still control them (Gandini, 2018; Kost et al., 2018; Gegenhuber et al., 2020).
15 The gig economy historically came about as a resurgence of early capitalistic precarious work:
on-call, piece-work compensation, home work, and triangular contracting arrangement (Stanford, 2017). This means that job design is an important policy for the platforms.
Attempting to define the boundaries of tasks performed through platforms is one of the core HRM activities. Rosenblat and Stark (2016) add to this that gig work is often characterized as offering freedom, flexibility and entrepreneurship, while actually crafting and limiting the tasks of gig workers in a shroud of information asymmetry. Kuhn (2016) adds to this that although this means that freelancers working in platforms are limited in their freedom, they are protected from debtors or client bullying through the designed working environment.
As mentioned before, practices are the most visible element of HRM activities, and are observable in the HRM process. The next section clarifies that part of HRM activities.
2.2.3. HRM process
Under HRM process the actually implemented HRM activities, how they are perceived, and who conducts them are intended (Bowen & Ostroff, 2004). The most important objective for platform HRM activities is to align the strategic focus of the firm with the context and the workers (Ostroff & Bowen, 2000). Although platforms do not actually employ the workers, they practice HRM on and with the gig workers in the shape of -mainly- recruitment, training, remuneration, appraisal, and firing, among other activities (Lee, et al., 2015; Jarrahl &
Sutherland, 2018; Meijerink & Keegan, 2019). However, how these activities come about are different in online labour platforms as opposed to more traditional work settings.
To understand this, academics often draw on the intended-implemented-perceived HRM concept (van Mierlo, Bondarouk & Sanders, 2018). Khilji and Wang (2006) showed that there is often a difference between the HRM as planned by top management, the implementation by line management, and the perception (in their research called satisfaction) by the workforce.
They also showed that this difference results in reduced organisational performance. This model highlights why it is so different to study HRM in the gig economy, further explained hereafter.
2.2.3.1.Intended HRM processes
First off, the intentions of higher management is complicated because online labour platforms have no formalized HRM department/manager or strategy, because that would imply an employment relationship (Schmidt, 2017; Meijerink & Keegan 2019). Furthermore, the most important difference is that there is no head of a HR department setting a strategy for the platform, but marketeers and programmers are in charge of developing the intended HRM activities (Lee, 2018). This means that the intended HRM activities are set by financially focused professionals, that may hold different ideals than HR professionals.
2.2.3.2.Actual HRM processes
Furthermore, there are no middle and line managers to implement the HR practices, but all
actors in the gig economy are responsible for conducting ‘actual’ HRM activities. Normally
the employees perceive HR and react accordingly, but now they are executives as well as
receivers of HRM. This is not necessarily a complication, because Trullen, Bos-Nehles and
Valverde (2020) show that implementing HRM is a dynamic process that requires several
actors to work on an implementation for it to be successful. But what does complicate matters
16 is that the classical line manager – employee relationship does not exist in the gig economy.
As said before, in the gig economy the HRM activities are executed by all involved actors.
Below an overview per main actor, what HRM activities they typically implement:
Platforms are besides connecting gig workers with clients also providing them with a way of communicating problems, or on the contrary praise about that connection through rating systems for the gig workers. This either leads to more gigs or to disconnection from the platform for the gig worker. Platforms can also offer incentives both to the clients and gig workers. Clients are given the option to tip the gig worker for his/her performance through the platform (Fiverr, 2020). While the platform can incentivize gig workers to work on certain times or in certain areas to fulfil the higher demands on those times and/or in those areas (Wood et al., 2018). Lastly the platform is also responsible for offering the gig workers a way of communicating with one another and the platform for feedback, learning, and communication of concerns. Learning is sometimes also offered through 4
thparties, to not insinuate training and development that an employer would offer (Ettlinger, 2017). In short the platforms thus participate in training, and remuneration, while facilitating appraisal, recruitment, and firing activities. But most of these activities are executed by an algorithm, that makes real-time decisions based on the data it is fed, and the rules it receives from the marketing officers and programmers (Meijerink & Keegan, 2019). The algorithmic management implemented by platforms often leads to negative perceptions among gig workers, given the impersonal nature (Shin & Park, 2019)
Gig workers as the human capital of the platforms are active participants platforming HRM processes too. They primarily work to the best of their abilities through platforms, but by doings so they interact with the algorithms that match them to clients, so that clients keep coming back the platform. If or when they come across clients who are not requesting or behaving by the rules of the platform, the workers can negatively appraise clients to make sure they are banned from the platform (D’Cruz & Noronha, 2018). Besides that many platforms work with referral schemes to attract more gig workers (Meijerink & Keegan, 2019), but without the participation of gig workers using those schemes no extra gig workers will come through that channel. Aside from bringing in new gig workers, the more experienced gig workers can also help and socialize with their colleagues over the online forum belonging to the platform. This way newer gig workers bring better quality work to the table sooner than when they would have to figure things out for themselves. So from the base practices, mentioned before, the gig workers actively participate in recruitment, training, and appraisal.
Clients are also active contributors to the HRM structure of platforms. Gegenhuber, Ellmer and Schüßler (2020) described in detail how clients review gig workers’ performance through the rating systems in place. This helps expel ill-performing gig workers from the platform, but also opens the gig workers up to liability outside of their control (e.g. on Uber a client wanted to get to the airport before a certain time, but that is not possible given the time it takes to drive there, but yet the client reviews the gig worker for not performing well (Wu et al., 2019)).
Besides appraisal the client also pays for the work delivered to them. This can go through the
platform, where the platform takes its cut in some way. Lastly, by requesting a task the client
brings demand onto the platform, helping in continuity and planning activities of the platform
(e.g. on the meal delivery platforms, gig workers need to be available around the local meal
times, because then many requests will logically flow in). Clients thus participate in
remuneration, appraisal, workforce planning, and firing activities.
17 Fourth parties, lastly, are connected to platforms to provide training to the gig workers, to heighten the work performance (Ettlinger, 2017). To avoid the employment relationship this goes via fourth parties, and is often not called training, but ‘tips and tricks’ (Uber, 2020), or
‘additional learning’ (Upwork, 2020).
2.2.3.3.Perceived HRM processes
When platform workers receive HRM it is often through an app, usually by an algorithm, which is received varyingly by the platform workers (Lee, Kusbit, Metsky & Dabbish, 2015). This means that the perceived HRM may also be more distant from the intended HRM activities because it is often delivered impersonal. The bigger the gap between intended and perceived HRM, the more HRM performance may be impacted (Piening, Baluch & Ridder, 2014).
All-in-all, there are enough factors that make the HRM seen on OLPs so complex. On some platforms the algorithm matches supply and demand, on others the clients approach the gig workers, or the other way around. Besides that, on some platforms the gig workers can set their own prices, but on others gig workers have to accept the prices set by the platform or the clients.
OLPs thus differ in content and process of HRM. This paper suggests that this can be traced back to the task characteristics, and proposes that platforms offering jobs with task- characteristics that are harder to come by, have to focus their HRM more on keeping the workers with those characteristics. On the other hand clients that request work from workers with more rare characteristics have to reward better too in order to obtain their service/product.
The next section explains that in more detail.
2.3. Resource dependence theory
Organisations make decisions to adapt to their changing environments all the time, in order to grow and ultimately survive (Malatesta & Smith, 2014). OLPs are based on cutting edge technologies and are in a turbulently changing environment. HRM decision-making just as well as other strategic decision-making occurs to respond to the environment (Tyskbo, 2019). This is what this research is mostly interested in. The Resource Dependence Theory (RDT) takes decision making based on the environment as the starting point for explaining the behaviour that firms display. Therefore the resource dependence theory developed by Pfeffer and Salancik (1978) can be used to explain why OLPs make different HRM decisions.
Perhaps the clearest description of the RDT is the book review by Reitz (1979). Reitz reviewed Pfeffer & Salancik’s work, and first introduced the theory, and summarized the process predicted by the theory as a cycle of three steps:
1. All organisations in the environment need resources to survive, and this leads to interdependence between organisations and actors in their environment;
The first thing to note is that this theory places the organisation in an open-system environment.
This means that an organisation is not a closed entity closed off from the outside world, but as
an interacting part of the environment, influencing and being influenced by its surroundings
(Nienhüser, 2008). An organisation rarely has all the resources it needs to perform all its
business functions already present internally. Therefore, organisations turn to other, external
actors (e.g. suppliers, potential workers) in their environment that can provide them with
resources they need, but do not possess themselves. Resources are not per se raw materials and
other assets. People, intellectual property and even affiliations can be seen as resources (Boyd,
1990). Because organisations trade resources to access them, dependencies amongst one
18 another occur. As such, to operate effectively, organisations are dependent on external parties that possess necessary resources. For this study dependency is referred to as the influence of external factors on organisational behaviour (Hillman, Withers & Collins, 2009).
2. This interdependence lead to uncertainty, because of this organisations start to look for ways to reduce these uncertainties;
Then the first thing that happens after establishing the interdependency is assessing the risks (and opportunities) to the business that this presents. Here, it is important to note that an organisation’s level of dependence on a resource (and thus the risk to losing access to that resource) is dependent on the resource’s importance, abundance, and ownership.
The importance, or criticality, of a resource (Salancik & Pfeffer, 1977; Nienhüser, 2008; Drees
& Heugens, 2013) is determined by how valuable the resource is to the performance of the organisation. As an example, the website of a platform that all clients and gig workers see and use is important for the success of a platform, but the quality of the jobs posted and the work delivered matter more. The availability of labour is thus a more important resource than the infrastructure of the website, even though that is valuable too. As such the dependence of an organisation on a resource is greater when it is more valuable/important to an organisation .The abundance refers to the availability of the resource (Stern, 1979). For example Uber allows anyone with a car to be an Uber driver, whereas professional freelance-platforms such as GigNow can only make use of consulting professionals, of which the availability is lower. As such the dependence of an organisation on a resource increases when the resource is more scarcely available. Lastly, the ownership is about who controls the access to the resource. The owner of the resource has power over the organisation in need of it, equal to that organisation’s dependence on the resource. Power is defined as “the capacity to influence other people, that it is conferred by the control of resources (positive and negative outcomes, rewards and costs, information, etc.) that are desired, valued or needed by others and which make them dependent upon the influencing agent for the satisfaction of their needs or reaching their goals” (Turner, pp. 2, 2005). Given that dependence on a resource increases the level of power by the owner of the resource, RDT places a strong emphasis on the study of power dynamics (Hillman, Withers & Colins, (2009).
Namely if a firm comes to the conclusion that the critical resource is not widely available, and owned by another organisations, that means there is an extremely uncertain position, where another organisation has power over the firm. Therefore the firm will start to look for ways to reduce this uncertainty and limit the power that external actors have over the firm’s activities.
3. To reduce uncertainty the organisations engage in activities such as forming coalitions, pooling resources, and/or other survival strategies.
Natural responses of firms are to attempt to resolve their dependence or power deficit, and increase their power over others. Activities to do so are to merge, acquire, form alliances with, and interlock with other firms that have some form of power over them (Drees & Heugens, 2013). Although strategies such as interlocking might reduce autonomy, they increase validity towards their environment (Hillman et al., 2009; Drees et al., 2013).
In the context of online labour platforms, a complicated power struggle can be seen between
the platforms, the gig workers, and the clients. This occurs since these three platform actors are
interdependent (Meijerink & Keegan, 2019).
19 Platforms exist because they can connect supply and demand of gig work better, faster, or cheaper than the actors in the environment could on their own (Schmidt, 2017). Losing access to either the gig workers or the clients will result in the demise of an OLP. Most often platforms do not lack gig workers, but they still need to treat them in a way that they will not join a competing platform. The access to timely and qualitative gig workers determines the attractiveness to clients, who will only use the platform if it fulfils their needs (Ettlinger, 2017).
As such, OLPs are dependent on gig workers and requestors that supply the platform with a key resource: labour.
Gig workers are dependent on the infrastructure a platform provides. Because of this the resource dependence theory advices gig workers to diversify into other fields to increase their power (Dill, 1981), but gig workers are often active in the gig economy because traditional labour markets have no place for them (Goods et al., 2019), sealing their dependence. As such, for generating an income, gig workers are dependent on OLPs that provide them with a key resource: access to work.
Clients can usually fulfil their needs outside the gig economy, by internalizing the resources they need (e.g. recruit creative talents or programmers), but this is more time and cost-intensive than acquiring goods/services through platforms. Platforms are convenient because they can deliver the short-term needs of clients timely (de Stefano, 2016). Here clients are however dependent on both platforms and workers. That is, without them, clients are not able to outsource work and therefore, are dependent on OLPs and their freelance workers to access a key resource: labour.
However, some of the traditional responses to dependence do not pan out in the platform economy. Because of this, platforms engage in HRM activities to deal with the power balance.
Through HRM activities platforms attempt to manage the supply of human resources and the demand for human resources. This is a different approach than the RDT prescribes in the following ways.
Firstly, gig workers are not organisations that can be merged or acquired by the platforms, and
the platforms also do not want to, as they would lose the freelancer-structure they need to keep
costs low. Secondly, gig workers may be searching for ways to (re)gain power, but as they are
individual contractors -and not a united workforce- they do not own the collective bargaining
capacity to change the status quo, even though unions and worker organisations have arisen
over the years (Johnston & Land-Kazlauskas, 2019). Platforms on the other hand can tweak
their algorithms and acquire other platforms to change the circumstances to their favour, and
increase their power (Kellogg, Valentine & Christin, 2020). Bowman (1979) adds to this that
determining how to use resources is also part of the power over resources. In the case of gig
workers they can decide to work via one platform, the other, or both, but that is the extent to
which they control the usage of their resource. Platforms have the power to deny gig workers
the access to their network, and depending on the offer of suitable gig workers, can do so as
they desire. Clients in that equation can visit several competing platforms to fulfil their need,
making them relatively powerful, unless they require a product/service that is not abundantly
available, but merely on a single platform.
20
Figure 1: Interdependencies visualized
2.4.Propositions
This research paper attempts to find out whether task characteristics determine who has the power and what that means for the HRM activities. This will be done by posing two propositions, as the OLP context is highly complex and exploring whether they hold any sway is on a conceptual level.
The power interrelationship thus depends on the criticality, abundance, and the ownership of a resource. This paper argues that the combination of task characteristics impacts the criticality, abundance, and ownership of gig work, and therefore which actor is most powerful. To illustrate, a platform offers tasks that are standardised, require a low skill level, is performed online, is short of duration, crowd tasked, and requires a low personal investment. This means that any gig worker with an internet connection could perform the task, so gig workers are in abundance, the platform controls access to the resource, and the criticality is assumed low, since it does not seem like a specialized task. Based on this reasoning, (granted it includes some assumptions) the Platform would look like the most powerful actor based on the task characteristics. Now if a platform would intermediate tasks that are unstandardised, requires a high skill level, is performed offline, is of a longer duration, is individually tasked, and requires a high personal investment. The abundance of gig workers capable of performing the task is drastically lower, and is also the owner of the resource. The criticality to the platform is assumed to be higher to the platform, so in this situation the gig worker is assumed to be the most powerful actor. This reasoning leads to the following proposition:
P1. The power relations between the platforms, the clients, and the platform workers depend on the different task characteristics performed by platform workers.
This paper argues that the HRM activities observed on the OLP reflect this power distribution.
When reviewing criticality, abundance and ownership of the labour resource, and the resulting
power distribution, it is expected that the HR philosophy of OLPs reflects this.
21 As an example, when the skills of gig workers are critical to the business operations, the HR philosophy is expected to be more preservative of those gig workers, and less controlling (Ashford et al., 2018). This is because finding other skilled gig workers can be a challenging process, so they do not want to lose their current workers. It is expected that appraisal, payment and incentive policies are implemented to motivate gig workers to remain with the platform, with practices such as loyalty bonusses, and reputation building through reviews and endorsements.
Whereas an abundance of labour is expected to result in a more controlling philosophy in which human resources are considered less precious and more as replaceable parts, so they do not mind a high turnover rate (Wood et al. 2018). This means that it is expected that strict performance management and job design policies are implemented, with strict rating/evaluation practices to root out under-performers, and clear communication of task requirements to ensure standard quality results.
In a third example, where the client is the most powerful actor it is expected that the philosophy of platforms is to use the gig workers as best as possible to suit the clients’ wishes, through job design and performance management practices, using practices such as job duration monitoring, client protection, gig-worker planning, client price-setting (Kuhn, & Maleki, 2017).
Because of the different task characteristics offered on platforms it is expected that the power balance is different, and therefore also the HRM. This leads to proposition 2a:
P2a. The HRM content on online labour platforms depends on the power relationships between the platforms, gig workers, and clients.
When the power relationship and the resulting HR philosophy are established, the consequential HRM processes would then also be clarified. Based on the power relationship that occurs the OLPs may choose to make other parties responsible for certain HRM activities.
As an example, when the client is the most powerful actor, it is to be assumed that the client sets the prices for the gigs, the clients gets to select who get the gig they are offering, and the clients get to determine the quality of the delivered work.
On the contrary, when the gig worker is the most powerful actor on the platform, it is expected that the gig worker can set their own price, determine which gig to take on, and are offered opportunities to develop their qualifications with 4
thparties.
Lastly, when the platform is the most powerful actor, it is expected that the platform regulates, the matching of gig worker and client, determines the payment that stands for the tasks, determine what requirements the task comprises, and who has to review who after the gig is completed.
So the HRM process and who performs what HRM activities leads to proposition 2b:
P2b. The HRM process on online labour platforms depends on the power relationships between the platforms, gig workers, and clients..
Below figure 2displays the conceptual model of the propositions:
22
Figure 2 Conceptual model for the propositions
